Biomachina Research History   by Willy Wriggers

As a graduate student and postdoc in the 1990s I was fascinated by the newly solved structures of G- and F-actin. I performed some of the earliest Molecular Dynamics (MD) simulations of actin and of the motor protein kinesin and collaborated with various experimental groups in the motility field. The work also led to significant method development, notably in protein domain motion analysis.

In 1998 I came across a research problem that dramatically accelerated my academic career. I developed a coarse-grained model and used it for bringing G-actin structures into register with F-actin electron microscopy (EM) maps that were rendered at the same level of detail [see paper]. This work led to the development of the award-winning Situs package, opened the door to a faculty position at The Scripps Research Institute, and it created opportunities for rigid body and flexible fitting of structural data from various biophysical origins.

The integration of multi-scale data from various biophysical origins soon became our signature research area and earned us an NIH R01 grant, due to the need in the "low resolution" community to produce atomic models for their data. Correlation techniques had arguably the biggest impact in this approach. In 2001 Pablo Chacón invented the Laplacian filter and implemented Fast Translational Matching in the now-famous colores program.  Julio Kovacs and Yao Cong later developed the Fast Rotational Matching approach in 2D, 3D, and 6D which earned us a HFSP project grant.

Haptic devices and virtual reality became a research focus after 2001. Force feedback during interactive EM fitting in our 3D theater was an amazing experience for the user. The focus in later years turned to improving the graphics and functional capabilities of the Sculptor program developed in the lab. Our work was far ahead of what much bigger groups and consortia were developing at the time.

During my work on actin in the 1990s I met the inventor of elastic network models, Monique Tirion. In 2002 we applied ENMs for the first time to EM maps in collaboration / competition with Florence Tama and Charles Brooks. Later, some of our ideas led to the ModeHunter package which is still actively developed. 

Large scale MD simulations produce an immense quantity of data. Principal Component Analysis (PCA) is a standard mathematical tool used to detect correlations in large data sets, but one problem we encountered in the 1990s was that the PCA modes suffer from the MD sampling problem. Zhiyong Zhang and I therefore developed a localized set of basis functions to discriminate relevant conformational changes in a protein from the background of atomic fluctuations. More recently, I have adapted computational geometry and network techniques to the study of long-time scale trajectories generated at D.E. Shaw Research.